11994
J. Phys. Chem. 1996, 100, 11994-12001
Femtosecond Infrared Studies of the Dissociation and Dynamics of Transition Metal Carbonyls in Solution Tianquan Lian, Steven E. Bromberg, Matthew C. Asplund, Haw Yang, and C. B. Harris* Department of Chemistry, UniVersity of California, Berkeley, California 94720, and Chemical Science DiVision, Lawrence Berkeley National Laboratory, Berkeley, California 94720 ReceiVed: June 18, 1996X
The ultrafast dynamics of the dissociation of M(CO)6 (M ) Cr, W, Mo) in alkane solutions were studied by femtosecond IR spectroscopy. After UV photolysis at 295 nm, both the bleach of the parent molecules and the absorption of the pentacarbonyl intermediate were probed with 240 fs time resolution. Oscillatory perturbed free induction decay signals before t ) 0 were observed and well characterized by realistic parameters of the system. The bleach recovery dynamics were found to be wavelength dependent, indicating that hot parent molecules are formed and that the bleach recovery time is determined by the vibrational cooling time. The measured percentage bleach recovery in n-heptane is less than the expected value calculated from the photosubstitution quantum yield measurements, suggesting that the initial recovery of the bleach is faster than our time resolution. The kinetics in the A1 vibrational mode region of the pentacarbonyl species have been measured to probe the formation and decay of the nascent product. The absorption of the product rises with an instrument response limited rate indicating that the formation of the product is much faster than 240 fs. The long time kinetics in this region reflect the vibrational cooling of the product. A fast decay with time constant of less than 300 fs is present in all the wavelengths probed, and its spectrum appears to resemble the early time spectrum of the hot pentacarbonyl species. This fast decay, observed in all three different metal carbonyls, is attributed to the fast geminate recombination of the pentacarbonyl with photodissociated CO ligand after only one to two collisions with the solvent cage.
Introduction Photolysis of transition metal carbonyls produces short-lived coordinatively unsaturated metal carbonyl intermediates, many of which catalyze important industrial or synthetic reactions.1-3 Knowledge of the initial formation, relaxation, and reactivity of these intermediates is therefore crucial to the understanding of the reaction mechanism. For this reason, the photochemistry of transition metal carbonyls has been extensively studied by ultrafast laser spectroscopy.4-13 In particular, with the advent of femtosecond infrared spectroscopy, a technique pioneered by Hochstrasser and co-workers,14 the detailed structure of shortlived reactive intermediates can be directly studied. The metal hexacarbonyls, M(CO)6 (M ) Cr, W, Mo), have long been used as model systems for the study of transition metal carbonyl photochemistry. Laser photolysis of gaseous M(CO)6 under collisionless conditions proceeds through the loss of several carbonyl groups.15 In a liquid solution16 and low-temperature matrix,17 however, UV photolysis leads to efficient loss of a single CO ligand to form a pentacarbonyl intermediate. The lack of multiple CO ligand loss was attributed to the efficient energy relaxation of the highly energized metal carbonyl intermediates in the condensed phase environment.10 Picosecond time-resolved experiments indicate that the naked pentacarbonyl species Cr(CO)5 is solvated within 1-2 ps in methanol7,8 and cyclohexane.7 The subsequent vibrational cooling of the solvated intermediate has been extensively studied by transient visible4,7,8 and IR6 absorption spectroscopy and resonant Raman spectroscopy.9 As a result of the photolysis, both the high-frequency CO stretching mode and other lowfrequency modes are excited. While the isolated high-frequency CO stretching modes relax on the 150 ps time scale,4,6,9 the * To whom correspondence should be addressed. X Abstract published in AdVance ACS Abstracts, July 15, 1996.
S0022-3654(96)01803-5 CCC: $12.00
low-frequency modes in the molecule relax more rapidly, taking tens of picoseconds.4,6 The time scale for the vibrational relaxation of these CO modes is similar to that of the parent molecules measured in IR pump and IR probe experiments.18-20 While the solvation and vibrational relaxation of the pentacarbonyl species are well characterized, the early time dynamics such as dissociation and cage recombination are not as well understood. UV/vis spectra of these molecules in alkanes show a strong singlet transition around 290 nm and a weaker triplet transition around 355 nm. It is not clear whether dissociation occurs from the singlet or triplet excited states. Wavelengthdependent photosubstitution quantum yield measurements on W(CO)6 in n-hexane solution have shown that the quantum yields are the same for excitation wavelengths ranging from 254 to 366 nm.21 This suggests that dissociation proceeds from the lower lying triplet state and that there is an effective intersystem crossing from the singlet state to the triplet state. Femtosecond spectroscopic studies on these molecules,8 however, suggest that dissociation may occur in the singlet state, as will be discussed in detail later. This model was also invoked in an earlier photoluminescence measurement in a low-temperature gas matrix.22 The photosubstitution quantum yields in various solvents have been measured, ranging from 0.42 to 0.93 depending on the solvent.23 These yields should be interpreted as the yield for formation of solvated W(CO)5 after the completion of any initial cage recombination. Two models have been proposed to account for the less than unit quantum yields.21,23 The first invokes a unit quantum yield for CO dissociation and a solvent-dependent, rapid geminate recombination of CO. The second model suggests that the quantum yield for CO dissociation itself is not unity and that efficient radiative and nonradiative relaxation channels exist for the excited state molecules. Although invoked to explain the nonunit quantum yield in © 1996 American Chemical Society
Transition Metal Carbonyls in Solution the photodissociation of M(CO)6 in solution,21,23 geminate recombination of CO has never been directly observed in these hexacarbonyls, nor has it been seen for other transition metal carbonyls. This process has, however, been observed in the solution phase photodissociation of simple diatomic molecules (I224 and I2- 25), polyatomic molecules: (CH2I226,27 and I3- 28,29), and metal carbonyl dimers.5,26 Cage recombination is not limited to solution phase reactions. It has also been observed in photodissociation reactions in solids,30 high-pressure gases,31 and clusters.32,33 To elucidate the role of geminate recombination dynamics in these metal carbonyls, ultrafast studies in the subpicosecond regime are needed. The ultrafast dynamics of M(CO)6 in alcohol and hexane solutions have been studied by femtosecond visible spectroscopy.8 A fast decay component with exponential time constant of 350 fs was observed in the 450 nm region after photolysis at 310 nm. This fast component was assigned to the absorption of an excited state while the 350 fs decay was interpreted as the time scale for the CO dissociation. This fast component was found to be similar for different central metal atoms, M ) Cr, Mo, and W, even though they have very different spin-orbit coupling. In light of this, it was suggested that the similarity in the early time dynamics indicated that either photodissociation occurs in the singlet state or the intersystem conversion is much faster than the 100 fs time resolution of the experiment. This fast component was observed in the same spectral region of the visible absorption of the M(CO)5 species. For this reason, it was later suggested to be a possible indication of a fast geminate recombination of the M(CO)5 fragment with CO.26 It is difficult to decide whether the absorption is due to M(CO)5 or the excited state of M(CO)6 by looking at the optical absorption spectra alone. The IR spectra of these two species are, however, expected to be quite different. Both ground state M(CO)6 and M(CO)5 have a strong and distinct IR absorption, making this reaction an ideal system for a femtosecond IR spectroscopic study.6 To unambiguously assign the previously observed fast component and understand the ultrafast dynamics in the solution phase photodissociation of M(CO)6, we have recently studied the bleach recovery dynamics of the parent CO stretching band and the product formation and decay dynamics for the three metal carbonyls in different alkanes using femtosecond infrared spectroscopy. Experimental Section The recently constructed femtosecond apparatus used in this study consists of a Ti:sapphire oscillator and dye amplifiers pumped by a 30 Hz Nd:YAG laser. The output of the Ti:sapphire oscillator34 is amplified in a two-stage LDS 798 dye amplifier and compressed using a pair of prisms to produce 10 µJ pulses with 50 fs pulse width at 820 nm. This amplified light is then split into three beams. One beam is further amplified to obtain 20 µJ pulses at 820 nm with pulse widths of approximately 70 fs. The other two beams are used to generate two separate sources of white light continuum in sapphire windows. Desired wavelengths of the white light are selected by two band-pass filters, each having a 10 nm fwhm window, and then amplified in two three-stage dye amplifiers to produce two independently tunable femtosecond pulses from 560 to 1000 nm. For this experiment, LDS 698 is used in one amplifier chain to obtain 100 µJ, 700 nm pulses. Rodamine 610 is used in another amplifier chain to produce 400 µJ, 590 nm light. The 590 nm light, which is compressed by a pair of prisms to shorten the pulse width to about 200 fs, is frequencydoubled in a 300 µm thick BBO crystal to generate UV light at 295 nm. IR pulses in the 5 µm region are generated by mixing
J. Phys. Chem., Vol. 100, No. 29, 1996 11995 the amplified pulse at 700 nm with the 820 nm pulse in a 2 mm LiIO3 crystal. The fwhm of the IR pulses is about 80 fs. The spectral band width is approximately 200 cm-1, limited by the phase matching condition of the LiIO3 crystal. By simply adjusting the phase matching angle, wavelengths from 1800 to 2300 cm-1 can be generated without changing band-pass filters. IR pulses in the 3-6 µm region can be generated by changing the band-pass filters and the dye for the amplifier. The important feature of this system is that it produces two independently tunable femtosecond pulses in addition to the 820 nm pulses, allowing us to generate independently tunable pump and probe pulses in UV, visible, and IR region. The 590 nm beam is sent through a variable delay line before being frequency doubled. The resulting UV beam, focused to a spot size of 200 µm at the sample, is used to initiate the photoreaction. The subsequent change in the IR absorption is probed by the IR beam. The probe IR beam is split into a signal and reference beam by a 50% Ge beam splitter. The signal beam, focused down to ca. 200 µM in diameter at the sample, is overlapped with the pump beam and recollimated afterward. Both the signal (after the sample) and reference beam are then focused into a monochromator (CVI, Digikrom 240). The two beams travel parallel to each other, displaced vertically by 1 cm along the direction of the monochromator slit. The slit size is variable from 10 µm to 2 mm. The IR beam size at the slit is about 300 µm, and the slit is set at 400 µm during normal operation. The two beams are separated after the monochromator and focused onto a matched pair of liquid N2 cooled HgCdTe detectors. The outputs from the two detectors are sent to two boxcar integrators and a digitizer. The high and low bounds from the signal and reference channels are set to be +40% and -40% of the average counts. The counts from the signal and the reference channels are normalized to remove shotto-shot laser fluctuations. The fluctuation of the IR pulse intensity is normally in the 10-40% range. This normalization scheme gets rid of the noise caused by the probe laser fluctuations. As a result, the noise of the system is dominated by the dark counts of the detectors. The typical noise of the system after averaging 100 laser shots is about 0.1%. The time dependence of a signal at a particular wavelength is collected by setting the monochromator at a fixed wavelength and changing the delay of the UV pump beam. A spectrum is recorded by scanning the monochromator while the pump pulse is fixed at a given delay time. The spectral resolution is determined by the resolution of the monochromator. For this experiment, a 150 groove/mm grating is used in the monochromator and the slit width is set at 400 µm, leading to a wavelength resolution of about 4 cm-1. The time resolution of the system is determined by the cross-correlation of the pump and probe beams, which is about 240 fs for this experiment. The instrument response can be measured in a thin silicon wafer. A UV photon excites the electron into the conduction band, giving rise to broad IR absorption. The rise time of the absorption signal, which is limited by the instrument response function, is well fitted by the integration of a Gaussian function with a fwhm of 240 fs. M(CO)6, obtained from Pressure Chemical, is dissolved in various alkane solutions. The sample is then pumped through a nozzle to produce a jet with a thickness of 200 µm in the center. For the experiments probing the bleach of parent molecules, the concentration of the sample, about 1 mM, is adjusted so that the OD of the CO stretching mode of M(CO)6 is ca. 1. For the experiments probing the M(CO)5 absorption, the concentration of the sample, at ca. 3 mM, is adjusted so
11996 J. Phys. Chem., Vol. 100, No. 29, 1996
Figure 1. Bleach recovery for M(CO)6 in heptane measured at the peak of the T1u CO stretching band at 1987, 1983, and 1989 cm-1 for M ) Cr, W, and Mo.
Lian et al.
Figure 3. Bleach recovery kinetics for W(CO)6 in heptane at 1983, 1980, and 1978 cm-1. The peak of the T1u CO stretching band is at 1983 cm-1.
Figure 2. Bleach recovery measured at the peak of the T1u CO stretching band for Cr(CO)6 in heptane and dodecane.
that the OD of the sample at 295 nm is ca. 1.0. The integrity of the samples is checked before and after experiments by FTIR spectra. Results Figure 1 shows the bleach recovery dynamics of M(CO)6 in heptane solution measured at the peak of the parent molecule T1u CO stretching mode. The squares, triangles, and circles are data points for M ) Cr (1987 cm-1), W (1983 cm-1), and Mo (1989 cm-1), respectively. The kinetics are fit by a constant bleach plus a single-exponential rise function with a rise time of 110, 70, and 70 ps as indicated by the thin lines. The percentage bleach recovery, calculated as the ratio of the amplitude of the exponential rise component to the total initial bleach, is 15%, 15%, and 6% for M ) Cr, W, and Mo, respectively. For comparison, the signals from different samples are scaled to have the same initial bleach value. Bleach recovery dynamics in different solvents are also measured. The bleach recovery dynamics for Cr(CO)6 in heptane and dodecane at 1987 and 1988 cm-1 are shown in Figure 2. The measured percentage bleach recovery is 15% and 21% in heptane and dodecane. These kinetics are measured at the peak of the parent molecule CO stretching band. As will be discussed later, these values for bleach recovery should not be interpreted as the amount of parent molecules re-formed, nor should the recovery time constant be taken simply as the time for re-forming the parent molecules. The recovery dynamics for all three M(CO)6 in different solvents are wavelength dependent near the parent molecule absorption band. Shown in Figure 3 are the kinetics
Figure 4. Perturbed free induction decay and population dynamics for Mo(CO)6 in heptane measured at the peak of the T1u CO stretching band at 1989 cm-1 and at 5, 10, and 20 cm-1 to the lower frequency side of the peak. The solid lines are fits to the data using the model described in the text. The broken line in the first panel is an integration of the instrument response function.
for W(CO)6 in heptane at 1983 cm-1 (circles), corresponding to the peak of the bleach, 1980 cm-1 (squares), and 1978 cm-1 (triangles). The data taken at 3 cm-1 on the high-frequency side is indistinguishable from that taken at the peak and is not shown for clarity. This wavelength-dependent behavior is a signature of the vibrational cooling of hot parent molecules, which will be discussed in next section. The spikes before t ) 0 seen in Figure 3 are due to perturbed free induction decay. We studied the detailed kinetics of the spikes to understand their influence on the early time dynamics. Shown in Figure 4 are the early time kinetics of Mo(CO)6 in heptane at its peak at 1987 cm-1 and at 5, 10, and 20 cm-1 on the lower frequency side of the peak. Similar kinetics for W(CO)6 in dodecane probed at 13, 23, 33, and 43 cm-1 to the low-frequency side of the parent molecule CO stretching band are shown in Figure 5. The solid lines are fits to the data using eq 1, which is explained below. The dotted line in the first panel in Figure 4 is the integration of the instrument response function, indicating the time evolution of an instantaneous bleach signal.
Transition Metal Carbonyls in Solution
Figure 5. Perturbed free induction decay and population dynamics for W(CO)6 in dodecane measured at 13, 23, 33, and 43 cm-1 to the lower frequency side of the peak of the T1u CO stretching band at 1982 cm-1. The solid lines are fits to the data using the model described in the text.
Figure 6. Kinetics of the formation of W(CO)5 in heptane probed at 1927 and 1917 cm-1, represented by the full triangle and square. The single-exponential fits, the dotted and solid lines, give time constants of 22 and 26 ps, respectively.
As shown in Figure 5, the perturbed free induction signal obscures the population dynamics around t ) 0 for probe wavelengths near the parent molecule bleach. To study the fast dynamics of the M(CO)5 species, we need to probe at a wavelength region as far away from the parent CO stretching band as possible. The vibrationally hot M(CO)5 species have broad absorptions, ranging from 1960 to 1900 cm-1.6 Shown in Figure 6 are the long time kinetics of W(CO)6 in heptane probing at 1927 cm-1 (the peak of the W(CO)5 A1 band) and 1917 cm-1. The kinetics can be well fit by a 22 ps exponential rise function at 1927 cm-1 and a 26 ps exponential decay at 1917 cm-1. These wavelength-dependent kinetics follow the vibrational cooling dynamics of the solvated W(CO)5X (X ) heptane) species.6 In addition to the cooling dynamics, there is also a fast initial decay in the kinetics when measured with small time steps of 33 fs. Shown in Figure 7 is the short time dynamics for Mo(CO)6 in heptane probed at 1929 (the peak of the Mo(CO)5 A1 band), 1918, and 1908 cm-1. The signals at different wavelengths are scaled to the same amplitude at the maximum absorbance change of the kinetics. The initial rise
J. Phys. Chem., Vol. 100, No. 29, 1996 11997
Figure 7. Early time dynamics for the formation and decay of Mo(CO)5 in heptane probed at 1929, 1918, and 1908 cm-1, represented by the full squares, triangles, and circles. The biexponential fits, the dotted and solid lines, give time constants of 150 fs for the fast component and about 20 ps for the slower component. The rise of the signal is instrument response limited.
of the signal is well fitted by an instrument response limited rise. All the kinetics are well fitted by a fast exponential decay followed by a slower exponential rise or decay function. The fast decay component has a time constant shorter than 300 fs. They appear to be fitted best by 100-200 fs exponential decays. Fits with 150 fs exponential decay are shown in Figure 7. The amplitude of the fast component cannot be well determined because it is very sensitive to the time constant used for the fit. Within the noise of the data, the relative amplitude for the fast and slow component appear to be similar for the three different wavelengths. The fast decay kinetics are also measured for different metal centers. Shown in Figure 8 are the kinetics measured at 1917 cm-1 for M(CO)6 in heptane with M ) Cr, W, and Mo, represented by square, triangle, and circles, respectively. The signals from different metal carbonyls are scaled to have the same size at the peak. The kinetics for different metals are similar, consisting of a fast component and a slower component with time constant of tens of picoseconds. The fast components are well fitted by exponential decays with time constants of about 150 fs, as shown in Figure 8. The relative amplitudes for the fast and slow components appear to be similar for different metal centers. Discussions Perturbed Free Induction Decay. Unlike electronic dephasing, which normally takes tens of femtoseconds,35,36 vibrational dephasing can be on the order of picoseconds. Vibrational coherent effects occurring on the time scale of dephasing will obscure the early time dynamics in a femtosecond IR spectroscopic study. A detailed study of these coherent effects is therefore necessary in order to understand the ultrafast chemical dynamics in the subpicosecond and picosecond time scale. Although these coherent effects have been well studied theoretically,37-40 there are very few detailed experimental studies using ultrafast infrared spectroscopy, where these effects can be more prominent due to the long dephasing time.41 The T1u mode of M(CO)6 in alkanes has a fwhm of about 3 cm-1. If one assumes homogeneous broadening, then this width corresponds to a vibrational dephasing time of about 1.8 ps.
11998 J. Phys. Chem., Vol. 100, No. 29, 1996
Lian et al.
Figure 8. Early time dynamics for the formation and decay of M(CO)5 in heptane probed at 1917 cm-1 for M ) Cr, W, and Mo, represented by the full squares, triangles, and circles, respectively. The biexponential fits, the dotted and solid lines, yield time constants of 150 fs for the fast component and about 20 ps for the slower component. The rises of the signals are instrument response limited.
Thus, coherent effects would be expected in our ultrafast study of these systems. The coherent effects in a pump probe experiment have been dealt with by many authors.37-41 Both pump-probe coupling and perturbed free induction need to be considered when the pulse width is shorter than the dephasing times. For a visible pump-probe experiment that involves all electronic transitions, both effects happen on the time scale of electronic dephasing time and need to be taken into account.39 For a visible/UV pump, IR probe experiment, the pump-probe coupling is important for the electronic dephasing time scale of tens of femtoseconds, while the perturbed free induction decay happens on the vibrational dephasing time scalestypically a few picoseconds. When the pulses are much longer than the electronic dephasing time but shorter or on the same time scale as the vibrational dephasing time, the pump-probe coupling terms can be neglected, and only the perturbed free induction decay terms are important. A detailed theoretical description of perturbed free induction signal in a visible pump and IR probe experiment has been worked out in a recent paper on the theory of vibrational spectroscopy.37 Upon absorption of a UV photon, M(CO)6 photodissociates to produce M(CO)5 and free CO. The oscillator strength of free CO is much smaller than bound CO and therefore can be neglected in our model.6 The rate of the formation of pentacarbonyl species is instrument response function limited, which we will discuss later. Therefore, we assume that the time of formation of M(CO)5 is shorter than the pump and probe pulse widths. The vibrational dephasing time is 1.8 ps in this system while the time resolution is about 240 fs. To simplify the problem without losing any physical insight, we assume that the probe pulse is much shorter than the dephasing time. The total signal from the system can then be described by the following formula:37
S(ω,τ) ) -C∑ j
{[
mj2
i∆j + γj
-
m12
] [ ] }
i∆1 + γ1 m12
i∆1 + γ1
θ(τ) +
m1mj
i∆j + γj
θ(-τ)e(i∆1+γ1)τ
-
(1)
where m1 is the oscillator strength of the parent molecule T1u
mode, mj the oscillator strength of the jth absorption band of M(CO)5, ∆ the detuning of the laser frequency from the center of the absorption bands, γ the homogeneous width of the bands, and θ(t) the Heaviside function. From eq 1, the perturbed free induction decay signal occurs before t ) 0 for a δ-function probe pulse. For a probe pulse that is not a δ-function, the perturbed free induction decay signal occurs before as well as during the pulse.37 One can still separate this coherent signal from the ultrafast population dynamics when the pulse is short compared with the time scale of the dynamics. The perturbed free induction decay signal has contributions from both the bleach and new absorption. While the bleach term depends on the oscillator strength of the bleach, the contribution from the new absorption term depends on the oscillator strength of both the bleach and the new absorption. The oscillation frequency of the perturbed free induction signal is determined by the detuning of the probe frequency from the bleach and the amplitude of the oscillation rises with a time constant corresponding to the dephasing time of the bleach. M(CO)5 has two absorption bands with a fwhm of a few wavenumbers when it is vibrationally cold. With UV excitation, however, this species is created with excess vibrational energy, and the early time spectrum is broad6 due to the excitation of the CO stretching mode and other low-frequency modes that are anharmonically coupled to it. The width of the W(CO)5 spectrum narrows in tens of picoseconds due to the cooling of low-frequency modes. The early time vibrational spectrum of W(CO)5 is thus modeled by a band centered at 1953 cm-1, the center of the E CO stretching mode, with a fwhm of 50 cm-1. To correctly describe the strength of the new absorption band, we allow the oscillator strength mj to vary so that it fits the long time signal, which is determined only by the relative contribution of absorption and bleach. The best fits to the data are shown by the solid lines in Figures 4 and 5. The parameters used for the fit are γ1 ) 5 cm-1, γ2 ) 50 cm-1, and m2/m1 from 1 to 2. The important features of the perturbed free induction decay signal are all reproduced by the fit, although there are two aspects of the model that can be further improved. Our fit is calculated at one single infinitely narrow frequency, while the experimental data are collected with a spectral window of about 4 cm-1. Better fits to the data can sometimes be obtained by changing the wavelength by a couple of wavenumbers, especially near the center of the bleach. A more sophisticated model should therefore include the effect of a finite spectral window. This is the likely reason that our data are better fitted with γ1 ) 5 cm-1 instead of 3 cm-1, the real width of the T1u mode measured with FTIR. Another improvement can be made by using the real time profile of the pump and probe pulses instead of the δ-function approximation used here, which is the main reason for the poor fit around t ) 0. We emphasize here that in a pump-probe experiment that detects only a small part of the probe frequency component, the observed bleach of a transition rises with the dephasing time of the transition instead of the instrument response function, as clearly shown in the first panel of Figure 4. Bleach Recovery. To understand the early time dynamics of the systems, we have probed the re-formation of the ground state parent molecules and the formation and decay of the product. The photosubstitution quantum yields for M(CO)6 in various alkane solvents have been previously observed to decrease as the chain length of the alkanes is increased. For Cr(CO)6 this value decreases from 0.72 in n-pentane to 0.58 in n-dodecane. Since the alkane-solvated pentacarbonyl species produced in the photolysis are completely substituted by the
Transition Metal Carbonyls in Solution entering ligands with a diffusion-limited rate in these reactions, the quantum yields measure the yields of producing the pentacarbonyl species after any initial cage recombination is complete. Therefore, the nonunit quantum yields suggest that after UV photolysis some of the parent molecules are regenerated in the ground state through either fast geminate recombination of the photofragments or radiative or nonradiative relaxation from the excited state. A previous ultrafast IR spectroscopic study on these molecules did not observe any absorption bands (from ca. 1 ps to ns time scale) other than that of the parent and the pentacarbonyl species,6 suggesting that it is unlikely that there is any excited state population on that time scale. The parent molecule would then be re-formed within 1 ps through fast cage recombination or nondissociative relaxation from the excited state, giving rise to a partial bleach recovery of the parent molecule absorption bands. A bleach recovery is indeed observed for all the three different metal carbonyls, as shown in Figure 1. The time scale for the bleach recovery cannot be taken as the time for re-forming the ground state molecules, but rather these dynamics reflect the vibrational cooling of the molecules. The presence of vibrational cooling of the ground state molecule is evident when we probe at the longer wavelength side of the bleach as shown in Figure 3. It has been shown5,6 that there are strong anharmonic couplings between the CO stretching and low-frequency modes for many metal carbonyls. As a result of populating the lowfrequency modes, the CO stretching bands are broadened and red-shifted. This effect was also observed for CO on metal surface.42 As shown in Figure 3, at the low-frequency side of the CO stretching band, the bleach recovers initially and then increases again. At the center and high-frequency side the bleach recovers monotonically with a slower time constant. The initial recovery in the kinetics at the low-frequency side is due to the absorption of the vibrationally hot parent molecules, which have a red-shifted CO stretching band, and the subsequent increase in bleach signal is caused by the vibrational cooling of those molecules. The recovery times measured at the center of the bleach are 112, 70, and 70 ps for M ) Cr, W, and Mo, respectively. These time constants reflect the cooling of both the low-frequency modes that are anharmonically coupled to CO stretching mode and the CO stretch itself. Unfortunately, the presence of vibrational population at V ) 1 and higher level in the CO stretching mode cannot be unambiguously determined in this experiment since their absorption overlaps with the absorption of the pentacarbonyl species. The vibrational cooling rates of low-frequency modes in these hexacarbonyls molecules have not been measured, but those for the pentacarbonyl species have been studied.6 It was found that the CO stretching bands of these pentacarbonyl species in hexane narrows in about 10 ps. This time constant is taken as the average cooling time of the low-frequency modes, which is similar to other large polyatomic molecules.43 The CO stretching time of M(CO)6 in n-hexane has been measured to be about 140 ps. Our measured bleach recovery times fall between 10 and 140 ps, suggesting that both low-frequency modes and CO stretching mode are excited when the molecule is re-formed in its electronic ground state. From the quantum yields for substitution in n-heptane of 73%, 79%, and 93% for Cr(CO)6, W(CO)6, and Mo(CO)6, the percentages of parent molecules re-formed are expected to be 27%, 21%, and 7%, respectively. The actual percentage of bleach recovery, as shown in Figure 1, is measured to be 15%, 15%, and 6% for the three metal carbonyls in n-heptane. The relative amounts of recovery agree with the trend expected from the substitution quantum yields, but the absolute values are
J. Phys. Chem., Vol. 100, No. 29, 1996 11999 significantly lower than those expected. This suggests that the re-formation of parent molecules is faster than we can measure at this wavelength. There should be a fast initial recovery of the bleach followed by the slower vibrational cooling shown in Figures 1 and 2. Unfortunately, as shown in Figure 4, the interference of the perturbed free induction decay signal makes it difficult to measure any fast dynamics in this spectral region. The bleach recovery for Cr(CO)6 in n-dodecane is also measured. Compared with n-heptane solution, shown in Figure 2, the relative amount of bleach recovery agrees with the trend that there is a higher percentage of recovery in a longer chain alkane solvent. The measured 21% recovery for Cr(CO)6 in n-dodecane is again smaller than the 42% expected from the quantum substitution yield measurement. Product Dynamics. Our measurements of the bleach recovery in these metal carbonyls suggests that there is a fast regeneration of the vibrationally hot ground state parent molecules. This recovery could be a fast relaxation from the electronic excited states and/or a fast cage recombination of CO and the pentacarbonyl intermediate. To understand the possible contribution of geminate recombination to the fast bleach recovery and therefore the nonunit quantum yield, we have studied the ultrafast dynamics of the pentacarbonyl species. As shown in Figures 4 and 5, to avoid the interference of the perturbed free induction decay signal, the measurements have to be made as far away from the parent molecule bleach as possible. At the peak, ca. 1927 cm-1, and lower frequency side of the A1 vibrational band of the three solvated M(CO)5, the contribution from the perturbed free induction decay can be neglected, and the product still has a significant absorption. The signal at all wavelengths can be fit by an instrument response time limited rise followed by a fast decay of ca. 150 fs and a slower wavelength-dependent decay or rise. Typical kinetics of the slow component are shown in Figure 6. The wavelength dependence and time constants for this component follow the vibrational cooling dynamics of the solvated pentacarbonyl species, which has previously been well studied by ultrafast visible4 and IR6 spectroscopy. At 1927 cm-1 (the peak of the A1 band), the slow component grows with an exponential rise time constant of about 22 ps, while at 1917 cm-1 (to the red side of the peak), the signal decays with a time constant of about 26 ps, reflecting the spectral narrowing and blue-shifting as a result of the cooling of the low-frequency modes that are coupled to the CO stretch. The fast decays appear to be faster than 300 fs at all wavelengths and are best fit by a time constant of approximately 150 fs. The detailed wavelength dependence for Mo(CO)6 in heptane is shown in Figure 7. The relative amplitude of the fast and slower components seems to be the same for all the three wavelengths, covering the peak of A1 band and 10 and 20 cm-1 away from this peak. Since the slower components are unambiguously assigned to the cooling of vibrationally hot pentacarbonyl species, the presence of the fast decay in the same spectral region with the same relative amplitude across the A1 band suggests that this component also pertains to the nascent pentacarbonyl species. The fast decay is then attributed to the rapid geminate recombination of the pentacarbonyl species with the CO ligand that is still trapped in the solvent cage. The 150 fs time scale for the geminate recombination agrees with previous measurements in CH2I2 and metal carbonyl dimer, Mn2(CO)10, in which geminate recombination was observed to happen on the 300 fs time scale.26 This ultrafast time scale for geminate recombination was also observed in a recent study on the dissociation of I2‚Arn clusters,32 in which it was found that geminate recombination of the I atoms occurs 360 fs after
12000 J. Phys. Chem., Vol. 100, No. 29, 1996 the direct dissociation of I2. Unfortunately, since the fast decay is faster than the time resolution, its amplitude cannot be accurately measured by the present experiment, and therefore we cannot determine quantitatively the percentage of pentacarbonyls that recombine. The kinetics measured at 1929 cm-1 for all three M(CO)6 (M ) Cr, and Mo) molecules are compared in Figure 8. The best fits to the fast components give time constants of approximately 150 fs for all three metal carbonyls. Again, neither the amplitude nor the time constant for the fast component can be accurately determined because of the lack of time resolution. If the quantum yield for CO dissociation is unity, we would expect to observe 27%, 21%, and 7% geminate recombination for Cr, W, and Mo, respectively, based on the measured photosubstitution quantum yield.23 If we observe similar percentages of recombination for all three different metal carbonyls, the different photosubstitution quantum yields must then be caused by different yields for CO photodissociation from these molecules. At this point, the quality of the data does not allow a quantitative comparison between the time scale and percentage of geminate recombination for the three different metal carbonyls. Therefore, the branching ratio between the dissociative and nondissociative pathway cannot be quantitatively determined. The instrument response limited rise of the signal indicates that the pentacarbonyl species are produced much faster than 240 fs. Based on this result, the previously observed instrument response limited rise of the visible absorption signal in the 450 nm region may be assigned to the formation of the pentacarbonyl species.8,26 Furthermore, the CO dissociation process is then complete within 100 fs.8 Unfortunately, the time resolution of about 250 fs in this experiment is not fast enough to provide an accurate measurement of the percentage of molecules that undergo geminate recombination. However, it is now possible to carry out femtosecond IR experiment with 40 fs time resolution.44 Future experiments with time resolution on the order of 40 fs should allow a quantitative measurement of the percentage and time scale of geminate recombination. A comparison between the three different metal centers should shed light on how the electronic properties of the metal center affect the dissociation dynamics of these metal carbonyls. A solvent dependence study should reveal what kind of solvent properties determine the time scale and amount of geminate recombination in the solution phase reaction. Conclusion UV photolysis leads to an efficient CO loss in M(CO)6. Both the bleach recovery and product formation and decay have been studied by femtosecond IR spectroscopy with 240 fs time resolution. The percentages of the bleach recoveries are found to be smaller than the photosubstitution quantum yields, indicating that re-forming of the vibrationally hot ground state parent molecules occurs faster than our time resolution. The relative amount of bleach recovery observed for the three metal carbonyls and in different alkanes agrees with the trend expected from photosubstitution quantum yield measurement. The formation of the pentacarbonyl species is found to be instrument response limited. We observe a fast decay (
